DE69029429T2 - Binary image reduction process - Google Patents

Description

BACKGROUND OF THE INVENTION (Field of invention)

The present invention relates to a binary image reduction method which is free from deterioration of image quality of an image represented by black and white, i.e., binary values, and more particularly to a binary image reduction method which can prevent thin lines from being lost, so that an uneven line generated when an image is read by an image reading apparatus such as a facsimile machine can be prevented from being separated.

(Description of the state of the art)

As a conventional binary image reduction method, an SPC (selective processing conversion) method shown in Fig. 1(a), a partial image mapping method shown in Fig. 1(b), and the like are known.

In the SPC method, when mapping a converted image onto an original image, a value of a pixel of the original image (hereinafter referred to as an original pixel) at a position closest to a pixel of the converted image to be obtained (hereinafter referred to as a converted pixel) is exchanged with the converted pixel. In Fig. 1(a), the value of an original pixel P0.0 is exchanged with a converted pixel Q. In the partial image mapping method, four original pixels near a converted pixel are selected; an average density projected by these four original pixels to the position of the converted pixel is calculated; and the calculated density is subjected to threshold processing to be converted into the value of the converted pixel. In Fig. 1(b), the value of a converted pixel Q is obtained based on four original pixels P0.0, P0.1, P1.0 and P1.1.

However, in the above reduction methods, if a black or white thin line is present on an original image, In the SPC method, when the thin line on the original image is not an original pixel closest to the position of the converted pixel, or in the partial image mapping method, when the result of thresholding an average density projected by four original pixels does not agree with the value of the thin line, the corresponding thin line is lost from a converted image, resulting in problems of jumping and buckling or defective typesetting. The "jumping" occurs when a black thin line of an original image is lost from a converted image, and the "bumping" occurs when a white thin line is lost. In Figs. 1(a) and 1(b), the value of the converted pixel Q indicates white, and a black thin line (P1.0, P1.1) of an original image is lost.

For a better understanding of the present invention, it should be added that, as a method for preventing jumping or buckling caused by a lost thin line, an improved TP method (thin line obtaining method) has been proposed by the applicant of the present application in Japanese Patent Application Publication No. 62-236087 filed on September 22, 1987 under JP-A-1 080 167, which document is not relevant prior art for the present application. In the TP method, a thin line is detected on an original image, and the detected thin line is obtained on a converted image, thereby preventing jumping or buckling. Fig. 2 shows the basic principle of obtaining a thin line. In Fig. 2, reference characters P0.0, P0.1, P1.0, P1.1, P2.0 and P2.1 denote original pixels; and Q denotes a converted pixel. Assuming that (Pm.m) represents a value of the original pixel,

represents an inverse value of (Pm.m) and (Q) represents a value of the converted pixel, then when (P0.0) = (P2.0) =

applies, these pixels are detected as a thin line and (Q) = (P1.0) is set.

When the TP method is applied to a reduction ratio of 1/3, a thin line is determined using original pixel references shown in Fig. 3. More specifically, six original pixels are referred to in both the horizontal and vertical directions (i.e., forming a cross shape) to have an original pixel closest to a target value of converted pixels as a central pixel. In Fig. 3, conditions for determining a thin line are as follows: when one of (P0.0) = (P-2.0) =

(P0.0) = (P2.0) = ( ) and (P0.0 = (P3.0) = ( ) = ( ), the presence of a vertical thin line is determined. If one of (P0.0) = (P0.-2) =

(P0.0) = (P0.2) = ( ) and (P0.0) = (P0.3) = ( ) = ( ), the presence of a horizontal thin line is determined. However, an image read by a scanner such as a facsimile machine often has an uneven pattern. In the TP method, when the thin line contains the uneven pattern, this pattern is recognized as a thin line in an opposite color and is retained, resulting in a loss of a thin line. Figs. 4(a) and 4(b) show a case where an original thin line is separated at a reduction ratio of 1/3 on a converted image because an original image contains an uneven pattern. More specifically, since the condition for determining a thin line of (P0.0) = (P0.2) =

formed, it is determined that a vertical white thin line is present at the position of an original pixel P1.0, and the value of a converted pixel Q in Fig. 4(a) is set to a white value. As a result, a converted image becomes as shown in Fig. 4(b).

Therefore, the present invention relates to a binary image reduction used in a facsimile machine or the like, and its object is to provide a binary image reduction method which can reduce an image read by a scanner having an uneven pattern without being provided with cracks or bumps, and which can reduce an image with high quality even when a reduction ratio is as small as, for example, 1/4.

SUMMARY OF THE INVENTION

According to the present invention as claimed, there is provided a binary image reduction method for reducing an original binary image, each pixel being represented by a binary value, with a conversion ratio αx in the horizontal direction being smaller than 1 and with a conversion ratio αy in the vertical direction being smaller than 1, to obtain a converted binary image, wherein

αx satisfies the condition 1/nx > αx ≥ 1/(nx + 1), and

αy satisfies the condition 1/ny > αy ≥ 1/(ny + 1),

where nx and ny are natural numbers,

the method comprising successively determining the binary value of each of the converted pixels by repeatedly performing the following steps:

a) obtaining (S1 in Fig. 6) reference numerals 1-2 in Fig. 5) a positional relationship between a given one of the converted pixels and the original pixels on the original image when the converted image is mapped onto the original image;

b) setting (S2 in Fig. 6) a rectangular area (reference numeral 11 in Fig. 7) having a side of 2(nx + 1) original pixels in the horizontal direction and a side of 2(ny + 1) original pixels in the vertical direction such that the given converted pixel is at the central position;

c) using some or all of the 4(nx + 1) (ny + 1) original pixels in the rectangular area, detecting (S3, S4 in Fig. 6) a line segment whose horizontal line width is equal to or smaller than nx pixels and whose vertical line width is equal to or smaller than ny pixels, and which has a length equal to or greater than a predetermined value; and

d) if such a line segment is present in the rectangular region and the converted pixel is closest to the detected line segment, allocating (S5 in Fig. 6, reference numeral 6 in Fig. 5) the binary value of the pixels of the detected line segment to the given converted pixel;

e) if no such line element is present in the rectangular region, determining the binary value of the given converted pixel by another method.

In a preferred embodiment, the step of detecting the presence/absence of a line segment includes detecting that a line segment is present in the original image,

(1) if in the original image a certain pixel with a certain value has a right and a left adjacent pixel with a different value from the certain pixel, and pixels with the same value as the certain pixel are connected by not less than a predetermined number in the vertical or oblique direction, or

(2) if in the original image pixels having the same value are connected in the horizontal direction, and pixels adjacent in the horizontal direction to the two ends of the horizontally connected pixels have a value different from the connected pixels, the number of horizontally connected pixels is not is less than 2 and not greater than nx, and pixels having the same value as the horizontally connected pixels are connected by not less than a predetermined number in the vertical or oblique direction.

A difference in a characterizing feature between the reduction method of the present invention and a conventional reduction method is that in the present invention, the length of a thin line is detected so as not to erroneously detect an uneven pattern as a thin line, and means for referring to original pixels having a size described in step (2) is used to allow detection of the length of a thin line.

As a result, according to the binary image reduction method of the present invention, even when a thin line contains an uneven pattern, a thin line can be prevented from being lost and a converted image free from deterioration can be obtained. In addition, a thin line in the horizontal, vertical or oblique direction can be prevented from being lost and a satisfactory converted image can be obtained.

Therefore, even if a small character is reduced, the reduced character can be read, and a character that is 30% smaller than the smallest character that can be reduced by a conventional method can be reduced.

BRIEF DESCRIPTION OF THE DRAWINGS

Figs. 1(a) and 1(b) show conventional binary image reduction methods, in which Fig. 1(a) shows an SPC (Selective Processing Conversion) method and Fig. 1(b) shows a partial image mapping method;

Fig. 2 shows the basic principle of maintaining a thin line in a conventional reduction process;

Fig. 3 shows references of an original pixel in a TP (Thin Line Preservation) method as a conventional reduction method;

Figs. 4(a) and 4(b) are views for demonstrating that when an original image contains an uneven pattern, the conventional TP method, in forming a converted image, erroneously detects the uneven pattern as a thin line;

Fig. 5 is a block diagram showing a binary image reduction function according to an embodiment of the present invention;

Fig. 6 is a flowchart showing a binary image reduction processing according to the present invention;

Figure 7 shows an example of original pixels referred to by the method of the present invention;

Figs. 8(a₁) to 8(b₃) show a detailed execution method of a thin line detection according to the present invention;

Figs. 9(a₁) to 9(d₃) show conditions in detection of a thin line used in the present invention;

Fig. 10(a) to 12(l) show positions of detection of thin lines on original images according to the present invention, wherein

Fig. 10(a) to 10(h) show detection positions of orthogonal/oblique thin lines having a line width of one pixel and a line length of three pixels when a conversion ratio α satisfies 1/1 > α ≥ 1/2,

Fig. 11(a) to 11(f) show detection positions of orthogonal/oblique thin lines with line widths of one and two pixels and a line length of three pixels when a conversion ratio satisfies α 1/2 > α ≥ 1/3, and

Fig. 12(a) to 12(l) show detection positions of orthogonal/oblique thin lines having line widths of one, two and three pixels and a line length of three pixels when a conversion ratio α satisfies 1/3 > α ≥ 1/4;

13(a) to 13(c) show converted images obtained by converting a character (a Chinese character) by conventional binary image reduction methods and the binary image reduction method of the present invention; and

Figs. 14(a) to 14(c) show converted images obtained by converting an original image having vertical, horizontal and oblique lines by a conventional binary image reduction method and the binary image reduction method of the present invention.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

Fig. 5 is a functional block diagram of an embodiment of the present invention. In Fig. 5, reference numeral 1 denotes an original image reading means for reading out an original image from a memory or the like; and 2, a converted pixel position calculating means for obtaining the positional relationship between converted pixels and original pixels when a converted image is mapped onto an original image. The calculating means 2 determines the positional relationship between converted pixels and original pixels on the basis of a conversion ratio. Reference numeral 3 denotes an original pixel reference reading means for reading out values of 2(nx + 1) (horizontal direction) 2(ny + 1) (vertical direction) original pixels to have a target value of converted pixels as a central pixel; 4 denotes means for detecting a vertical/oblique thin line having a line width of nx pixels or less and a horizontal/oblique thin line having a line width of ny pixels or less from the values of the original pixel reference read out by the original pixel reference reading means 3 and for obtaining a converted pixel closest to the detected thin line; 5 denotes means for determining a length of a thin line for determining whether a thin line has a predetermined length or not; 6 denotes means for when a target value of converted pixels is a converted pixel closest to a thin line and the thin line has the predetermined length, exchanging the value of the thin line with the converted pixel; 7 denotes means for calculating a value of a converted pixel when no thin line is detected, or when the detected thin line has a length shorter than the predetermined length, or when a converted pixel closest to the thin line is not a target value of converted pixels; and 8 denotes means for determining a value of a converted pixel.

Fig. 6 shows a flow of binary image reduction processing according to the present invention. In the binary image reduction processing, values of converted pixels are sequentially determined according to a predetermined rule, and the determination operation is repeated to form a converted image. The embodiment shown in Fig. 5 and the binary image reduction method of the present invention will be described below with reference to the flow chart of Fig. 6.

When horizontal and vertical conversion ratios are different, conversion may be performed separately for the horizontal and vertical directions. Therefore, in the following description, it is assumed that the horizontal and vertical conversion ratios are equal to each other and a conversion ratio α;

1/n > α ≥ 1/(n+ 1)

satisfies, where (n is a natural number).

An original image is read out by the original image reading means 1, and a converted image is mapped onto the original image by the converted pixel position calculating means 2, thereby obtaining the position of a target value of converted pixels on the original image (S1 in Fig. 6).

Some or all of a total of 4(n+1)² (namely 2(n+1) in horizontal direction and 2(n+1) in vertical direction) original pixels are used as source pixel reference read out by the original pixel reference reading device 3 (S2 in Fig. 6).

The value of the natural number n then takes on the following values when the 4(n+1)² original pixels are read out, depending on the conversion ratioα:

If 1/1 > α ≥ 1/2 then n = 1

If 1/2 > α ≥ 1/3 then n = 2

If 1/3 > α ≥ 1/4 then n = 3

Fig. 7 exemplifies a case where the conversion ratio satisfies 1/3 > α ≥ 1/4. In Fig. 7, original pixels to be referred to are surrounded by a dashed line 11. Each character "0" indicates an original pixel, and a character "x" indicates the position of a target value of converted pixels. In this case, since n = 3, 64 (= 2(n+1) x 2(n+1) = 8 x 8) original pixels are referred to. In a detailed referencing method, the position of the target value of converted pixels on the original image is detected, and 2(n+1) (horizontal direction) x 2(n+1) (vertical direction) original pixels are sequentially read out from the contents of a line memory having a capacity of 2(n+1) lines to store an original image to have the target value of a converted pixel as a central pixel.

The thin line detecting and converted pixel value calculating means 4 detects connected pixels that are vertical/oblique connected pixels having a horizontal width of nx pixels or less and/or horizontal/oblique connected pixels having a vertical width of ny pixels or less, and the next converted pixel of these is the target value of converted pixels (S3 in Fig. 6).

When the connected pixels of which the next converted pixel is the target value of converted pixels are detected, the thin line length determining means 5 determines whether the connected pixels have a predetermined length or not (S4 in Fig. 6).

Figs. 8(a₁) to 8(b₄) show a detailed execution method of a thin line detection of the present invention. Figs. 8(a₁) to 8(b₄) show by way of example, vertical and horizontal thin lines having line widths of one and two original pixels and a length of three pixels when the conversion ratio α = 1/3. Fig. 8(a₁) shows an area defining thin line detection of a line width of one pixel, and Fig. 8(a₂) shows an area defining thin line detection of a line width of two pixels. Fig. 8(b₁) shows a thin line detection condition (hereinafter referred to as a window) for detecting a vertical line width of one pixel, Fig. 8(b₂) shows a window for detecting a vertical line width of two pixels, Fig. 8(b₃) shows a window for detecting a horizontal line width of one pixel, and Fig. 8(b₄) shows a window for detecting a horizontal line width of two pixels.

Thin lines are detected using the detection methods shown in Figs. 8(b₁) to 8(b₄) by sequentially detecting the presence/absence of thin lines in all areas of an original pixel reference shown in Fig. 8(a₁) or 8(a₂).

A detection condition of the window of a vertical line width of one pixel in Fig. 8(b₁) is given by:

(R-1.0) = (R1.0) = ( )

=

= ( )

A detection condition of the window of a vertical line width of two pixels in Fig. 8(b2) is given by:

(R-1.0) = (R2.0) = ( ) = ( )

=

+

= ( ) + ( )

where the sign "+" indicates a logical sum (OR).

A detection condition of the window of a horizontal line width of one pixel in Fig. 8(b₃) is given by:

(R0.-1) = (R0.1) = ( )

=

= ( )

A detection condition of the window of a horizontal line width of two pixels in Fig. 8(b₄) is given by:

(R0.-1) = (R0.2) = ( ) = ( )

=

+

= ( ) + ( )

Figs. 8(a₁) to 8(b₄) show some examples of thin line detection. Windows used in thin line detection may have different formats depending on the line widths and lengths of the thin lines and on a case of detecting only orthogonal thin lines or detecting both orthogonal and oblique thin lines.

9(a₁) to 9(d₃) show effective windows for detecting thin lines for practical use. When it is clear that a thin line having a line width of in pixels exists when m (m is a natural number) original pixels having the same value are connected, and original pixels adjacent to two ends of the m original pixels have a value different from those of the m original pixels, Figs. 9(a₁), 9(a₂) and 9(a₃) show windows for determining the presence of a thin line when a thin line having a line width of in pixels (where m = 1, 2 or 3) exists and the thin line of the m pixels has a length of two pixels or more in the vertical and horizontal directions; Figs. 9(b₁), 9(b₂) and 9(b₃) show windows for determining the presence of a thin line when a thin line having a line width of m pixels is present and the thin line of the m pixels has a length of two pixels or more in vertical, horizontal and oblique directions; Figs. 9(c₁), 9(c₂) and 9(c₃) show windows for determining the presence of a thin line when a thin line having a line width of m pixels is present and the thin line of the m pixels has a length of three pixels or more in vertical and horizontal directions; and Figs. 9(d₁), 9(d₂) and 9(d₃) show windows for determining the presence of a thin line when a thin line having a line width of m pixels is present and the thin line of the m pixels has a Has a length of three pixels or more in the vertical, horizontal and diagonal directions.

The windows shown in Figs. 8(b₁) and 8(b₃) exemplify a case of Fig. 9(c₁), and the windows shown in Figs. 8(b₂) and 8(b₄) exemplify a case of Fig. 9(c₂).

As described above with reference to Figs. 8(a1) to 8(b4), original pixels in a region defined by the conversion ratio are read out to have a target value of converted pixels as a central pixel for detecting a thin line having a line width determined by the conversion ratio or less, and only a thin line having a length longer than a reference value is retained. For this purpose, one of the windows shown in Figs. 9(a1) to 9(d3) is selected according to a necessary condition, and the presence/absence of a thin line satisfying the window is detected from the readout region. According to the method given above, thin lines are not lost.

In the method given above, the entire area of an original pixel reference read out to have the target value of converted pixels as a central pixel is checked to detect the presence/absence of a thin line corresponding to a window, and when a thin line is detected, it is checked whether a converted pixel closest to the thin line is the target value of converted pixels.

When the position of the target value of converted pixels is determined, the position of a thin line detection (hereinafter referred to as a window position) is determined according to conditions of the conversion ratio, the length and width of the thin line, and the direction in which the thin line extends (i.e., the vertical, horizontal, or oblique direction). Therefore, the presence/absence of a thin line can be detected from the window position determined as described above. Note that four different window positions are present according to a positional relationship between the target value of converted pixels and four original pixels surrounding the target value of converted pixels.

Figs. 10(a) to 10(h) show window positions used for detecting orthogonal/oblique thin lines having a line width of one pixel and a line length of three pixels, i.e., window positions that assume the window shown in Fig. 9(d1) when the conversion ratio α satisfies 1/1 > α ≥ 1/2. In Figs. 10(a) to 10(h), Q0 denotes a target value of converted pixels; and Pm.n (where m = -2 to 1 and n = -1 to 2) denotes original pixels. A part 21 surrounded by a dashed line corresponds to a window position. Figures 10(a), 10(c), 10(e) and 10(g) show window positions for vertical/oblique thin lines with a line width of one pixel, and Figures 10(b), 10(d), 10(f) and 10(h) show window positions for horizontal/oblique thin lines with a line width of one pixel.

Figs. 10(a) to 10(h) show four different cases by way of example, i.e., a case where the position of the target value Q0 of converted pixels is near the original pixel P0.0 (Figs. 10(a) and 10(b)); a case where the position of the target value Q0 is near the original pixel P0.1 (Figs. 10(e) and 10(f)); and a case where the position of the target value Q0 is near the original pixel P--1.1 (Figs. 10(g) and 10(h)). The four different positions of the target value Q0 of converted pixels correspond to positions rotated by 90°.

In some cases, the target value Q0 of converted pixels may be present at a position that is an equal distance from two adjacent pixels of surrounding four original pixels, or it may be present at a position that is an equal distance from all four surrounding original pixels. In these cases, the nearest original pixel is determined in advance so that the four different positions of the target value of converted pixels can be obtained.

Figs. 11(a) to 11(f) show window positions used for detecting orthogonal/oblique thin lines having line widths of one and two pixels and a line length of three pixels, that is, window positions that assume the windows shown in Figs. 9(d1) and 9(d2) when the conversion ratio satisfies α 1/2 > α ≥ 1/3. Note that Figs. 11(a) to 11(f) only exemplify a case where the target value Q0 of converted pixels is closest to the original pixel P0.0. Of course, the target value Q0 assumes four different positions as in Figs. 10(a) to 10(f). Figs. 11(a) and 11(c) show window positions for vertical/oblique thin lines having a line width of one pixel; Figs. 11(b) and 11(d) show window positions for horizontal/oblique thin lines with a line width of one pixel; Fig. 11(e) shows a window position for vertical/oblique thin lines with a line width of two pixels; and Fig. 11(f) shows a window position for horizontal/oblique thin lines with a line width of two pixels.

Figs. 12(a) to 12(l) show window positions used for detecting orthogonal/oblique thin lines having line widths of one, two and three pixels and a line length of three pixels, i.e., window positions that assume those in Figs. 9(d1), 9(d2) and 9(d3) when the conversion ratio α satisfies 1/3 > α ≥ 1/4. Note that Figs. 12(a) to 12(l) as well as Figs. 11(a) to 11(f) only exemplify a case where the target value Q0 of converted pixels is closest to the original pixel P0.0. Of course, the target value Q0 assumes the same as Figs. 11(a) to 11(f). as in Figs. 10(a) to 10(f). Figs. 12(a), 12(c) and 12(e) show window positions for vertical/oblique thin lines with a line width of one pixel; Figs. 12(b), 12(d) and 12(f) show window positions for horizontal/oblique thin lines with a line width of one pixel; Figs. 12(g) and 12(i) show window positions for vertical/oblique thin lines with a line width of two pixels; Figs. 12(h) and 12(j) show window positions for horizontal/oblique thin lines with a line width of two pixels; Fig. 12(k) shows a window position for vertical/oblique thin lines with a line width of three pixels; and Fig. 12(l) shows a window position for horizontal/oblique thin lines with a line width of three pixels.

All examples of the window positions shown in Figs. 10(a) to 12(l) are used for a case where the thin line has a length of three pixels. The window position for a thin line having a length of two pixels can be determined in the same manner as described above. The windows for thin lines having a length of two pixels are as shown in Figs. 9(a₁), 9(a₂), 9(a₃), 9(b₁), 9(b₂), and 9(b₃).

When a thin line having a predetermined length is detected by the thin line length determining means 5 shown in Fig. 5, the converted pixel value exchanging means 6 exchanges the thin line value with a value of the target value of converted pixels (S5 in Fig. 6).

When no thin line is detected, or when even when a thin line is detected, a converted pixel closest to the thin line is not the target value of converted pixels, or when the detected thin line does not have a predetermined length, the converted pixel value calculating means 7 shown in Fig. 5 calculates the value of the converted pixel (e.g., a value of the nearest original pixel) according to predetermined logical calculations, and exchanges the calculated value with the target value of converted pixels.

In this embodiment, a method of detecting that the target value of converted pixels of a thin line is closer than other converted pixels can be realized by calculating the central line of the detected thin line, comparing distances from the central line to adjacent converted pixels to detect the next converted pixel, and checking whether the next pixel is the target value of converted pixels. In another method, when a distance between original pixels is given by 1, a distance between converted pixels is given by 1/α when the conversion ratio is α. The center between two converted pixels is away from the two converted pixels by 1/2α. Therefore, if it can be detected whether the central line of a thin line is included in a region that is away from the target value of converted pixels by a distance of 1/2α, the distance between the converted pixels is 1/2α. in the left, right, upper and lower directions, check whether the converted pixel closest to the thin line is the target value of converted pixels.

In this embodiment, every 2(n+1) (horizontal direction) 2(n+1) (vertical direction) original pixels are read out, and a thin line is detected using some or all of the read out original pixels, thereby forming a converted image. However, a converted image may be formed by a method of reading out only original pixels necessary for detection of a thin line.

In the above description, the cases where thin lines with a length of two pixels or more are detected and where thin lines with a length of three pixels or more are detected are described. When comparing these cases, the following characteristics are found.

Assume that the case where thin lines having a length of two or more pixels are detected is hereinafter called method (1), and the case where thin lines having a length of three or more pixels are detected is hereinafter called method (2).

The number of pixels required for detection of a thin line of the method (1) shown in Figs. 9(a1) to 9(b3) is smaller than that of the method (2) shown in Figs. 9(c1) to 9(d3). For this reason, when detection of a thin line of the method (1) is realized in hardware, the number of hardware components can be smaller than that of the method (2); when realized in software, the program size can be smaller than that of the method (2), resulting in a higher processing speed. However, since thin lines having a length of two or more pixels are detected, when an uneven pattern of one pixel is included in a thin line, the thin line often cannot be prevented from being separated. For this reason, the image quality of a converted image using method (1) is inferior to that using method (2).

On the other hand, the number of pixels required for detection of a thin line in the method (2) is larger than that in the method (1). For this reason, when detection of a thin line in the method (2) is realized in hardware, the number of hardware components is larger than that in the method (1); when realized in software, the program size is larger and a processing speed is lower. However, in the method (2), since thin lines having a length of three or more pixels are detected even when an uneven pattern of one pixel is present in a thin line, the thin line is not separated. For this reason, the image quality of a converted image is superior to that in the method (1). Even when an uneven pattern of one pixel is present in a thin line, since the uneven pattern does not have a length of three or more pixels, the uneven pattern is never mistakenly detected as a thin line.

The above features can provide the following advantages. When the number of hardware components is to be reduced or when a higher processing speed is to be achieved, the method (1) is preferably used. When the image quality is more important, the method (2) is preferable. In this case, In this way, a desired method can be selected from the two methods.

Fig. 13(b) shows characters obtained by reducing an original image shown in Fig. 13(a) by conventional reduction methods (the SPC method and the TP method), and Fig. 13(c) shows a character obtained by converting the original image of Fig. 13(a) by the reduction method of the present invention. Fig. 14(b) shows a pattern obtained by reducing an original image shown in Fig. 14(a) having vertical, horizontal and oblique lines by the conventional reduction method (the SPC method), and Fig. 14(c) shows a pattern obtained by converting the original image of Fig. 14(a) by the reduction method of the present invention. As can be seen from these figures, binary image reduction without loss of thin lines can be realized by the reduction method of the present invention.

Claims (4)

1. A binary image reduction method for reducing an original binary image, each pixel being represented by a binary value, with a conversion ratio αx in the horizontal direction that is less than 1, and with a conversion ratio αy in the vertical direction that is less than 1, to obtain a converted binary image, wherein αx satisfies the condition 1/nx > αx ≥ 1/(nx + 1), and αy satisfies the condition 1/ny > αy ≥ 1/(ny + 1), where nx and ny are natural numbers, the method comprising sequentially determining the binary values of each of the converted pixels by repeatedly performing the following steps: a) obtaining (S1 in Fig. 6, reference numerals 1-2 in Fig. 5) a positional relationship between a given one of the converted pixels and the original pixels on the original image when the converted image is mapped onto the original image; b) setting (S2 in Fig. 6) a rectangular area (reference numeral 11 in Fig. 7) having a side of 2(nx + 1) original pixels in the horizontal direction and a side of 2(ny + 1) original pixels in the vertical direction such that the given converted pixel is at the central position; c) using some or all of the 4(nx+1) (ny+1) original pixels in the rectangular area, detecting (S3, S4 in Fig. 6) a line segment whose horizontal line width is equal to or less than nx pixels and whose vertical line width is equal to or less than ny pixels and which has a length equal to or greater than a predetermined value; and d) if such a line segment is present in the rectangular area and the converted pixel is closest to the detected line segment, allocating (S5 in Fig. 6, reference numeral 6 in Fig. 5) the binary value of the pixels of the detected line segment to the given converted pixel; e) if no such line element is present in the rectangular region, determining the binary value of the given converted pixel by another method. 2. A binary image reduction method according to claim 1, wherein the step of detecting the presence/absence of a line segment includes detecting that a line segment is present in the original image, (1) if in the original image a certain pixel with a certain value has a right and a left adjacent pixel with a different value from the certain pixel, and pixels with the same value as the certain pixel are connected by not less than a predetermined number in the vertical or oblique direction, or (2) when, in the original image, pixels having the same value are connected in the horizontal direction, and pixels adjacent in the horizontal direction to the two ends of the horizontally connected pixels have a value different from the connected pixels, the number of the horizontally connected pixels is not less than 2 and not more than nx, and pixels having the same value as the horizontally connected pixels are connected by not less than a predetermined number in the vertical or oblique direction. 3. A binary image reduction method according to claim 1, wherein the step of detecting the presence/absence of a line segment includes detecting that the line segment is present in the original image, (1) if in the original image a certain pixel with a certain value has an upper and a lower neighboring pixel with a different value from the certain pixel, and pixels with the same value as the certain pixel are connected by not less than a predetermined number in the horizontal or oblique direction, or (2) when in the original image, pixels having the same value are connected in the vertical direction, and pixels adjacent in the vertical direction to the two ends of the vertically connected pixels have a value different from the connected pixels, the number of the vertically connected pixels is not less than 2 and not more than ny, and pixels having the same value as the vertically connected pixels are connected by not less than a predetermined number in the horizontal or oblique direction. 4. A binary image reduction method according to claim 2 or 3, wherein the predetermined number corresponds to at least two pixels on the original image.